Submission · qwen3-8b

`fc88002bf2374bc2`

by NVFP4 GPTQ + scale search (reference) · test split · runner replay

NVFP4 GPTQ + scale search Qwen3-8B-NVFP4-chat-ss W4A4KV16

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

0.114LowBScore

accuracy gate: the worst domain's 95% upper bound on the drop is 5.57 points, over 5

W·A·KV	W4A4KV16
Accuracy	fail · mean drop 2.07 points
Speed gain	1.39× BF16 in the same pod (serve 1.33×, interactive 1.37×, long context 1.48×) (gate: at least 0.95× less its error)
Memory gain	1.42× BF16's peak (serve 1.22×, interactive 2.23×, long context 1.06×) · 45.7 GiB against BF16's 55.6 GiB on the first workload
Efficiency gain E	1.39
Per-domain results on the hidden split are not released.

Raw record

{
  "agent": "zoo-imx-gptq-nvfp4-uc",
  "created_at": "2026-10-07T07:22:40Z",
  "error": null,
  "id": "fc88002bf2374bc2",
  "participant": "NVFP4 GPTQ + scale search",
  "partition": "reference",
  "result": {
    "admitted": false,
    "bf16_peak_gib": 55.628,
    "drift": 2.0262,
    "efficiency_gain": 1.3913,
    "length_shift": 0.02358,
    "mean_drop": 0.02072,
    "memories": {
      "interactive": 2.2254,
      "long_context": 1.0626,
      "serve": 1.2172
    },
    "memory": 1.4225,
    "memory_saved": 0.1784,
    "metric": "lowbit_score",
    "peak_gib": 45.702,
    "precision": {
      "activations": 4,
      "kv": 16,
      "label": "W4A4KV16",
      "weights": 4
    },
    "qualified": false,
    "quality": {
      "tier": "fail",
      "valid_domains": 5
    },
    "reason": "accuracy gate: the worst domain\u0027s 95% upper bound on the drop is 5.57 points, over 5",
    "score": 0.1136,
    "score_se": 0.0036,
    "speed": 1.393,
    "speed_serve": 1.3309,
    "speeds": {
      "interactive": 1.3727,
      "long_context": 1.4795,
      "serve": 1.3309
    },
    "task_id": "qwen3-8b"
  },
  "runner": "replay",
  "solution": {
    "checkpoint": "Qwen3-8B-NVFP4-chat-ss",
    "declared_precision": false,
    "detected_precision": "W4A4KV16",
    "method": "NVFP4 GPTQ + scale search",
    "precision": "W4A4KV16"
  },
  "split": "test",
  "stages": [
    {
      "name": "build",
      "status": "done"
    },
    {
      "name": "contract",
      "status": "done"
    },
    {
      "name": "quality",
      "status": "done"
    },
    {
      "name": "efficiency",
      "status": "done"
    },
    {
      "name": "score",
      "status": "done"
    }
  ],
  "status": "done",
  "task_id": "qwen3-8b",
  "track": "reference",
  "updated_at": "2026-10-09T04:57:10Z"
}